You are on page 1of 10

Mathematica Slovaca

Djurdje Cvijovi; Jacek Klinowski


Finite cotangent sums and the Riemann zeta function
Mathematica Slovaca, Vol. 50 (2000), No. 2, 149--157

Persistent URL: http://dml.cz/dmlcz/131809

Terms of use:
Mathematical Institute of the Slovak Academy of Sciences, 2000
Institute of Mathematics of the Academy of Sciences of the Czech Republic provides access to
digitized documents strictly for personal use. Each copy of any part of this document must contain
these Terms of use.
This paper has been digitized, optimized for electronic delivery and stamped
with digital signature within the project DML-CZ: The Czech Digital Mathematics
Library http://project.dml.cz

Mathematica
Slovaca
2000

. . . _.
__, /^rt-.rtx M ^ ->!/-. ni-Math. SlOVaCa, 5 0 ( 2 0 0 0 ) , NO. 2, 1 4 9 - 1 5 7

Mathematical Institute
Slovk Academy of Sciences

F I N I T E C O T A N G E N T SUMS A N D
T H E R I E M A N N ZETA F U N C T I O N
D J U R D J E CVIJOVIC

JACEK KLINOWSKI

(Communicated by Stanislav

Jakubec )

A B S T R A C T . We give a closed-form evaluation of t h e finite cotangent sums given


by

- ('D

[ffitfc].
L

p=0

p=l

where n and q are positive integers {q > 2) and is a non-integer real number.
The latter sum enables us to simplify t h e Apostol formula concerning t h e R i e m a n n
zeta function at integer values of t h e argument. We d e m o n s t r a t e t h a t , for even
integers, this formula involves polynomials with rational arguments.

1. Introduction
Consider the Riemann zeta function ((z) defined by [1; p. 19]
oo

..

CW =

=
F

oo

Rez>1

T^S(^TF

k=l

k=l
oo

v
1

= r_i--/E(- 1 )' k "V

'

Rez>0

k=i

' z*1-

For Re z < 1, z ^ 1, ((z) is defined as the analytic continuation of the foregoing


series. The Riemann zeta function is analytic over the whole complex plane,
except at z 1, where it has a simple pole.
A p o s t o l [2] gave the following formula

cot" (-^r)
C(n) - ( J )
v

V2j

lim
?->oo

qn

(1)
v

1991 M a t h e m a t i c s S u b j e c t C l a s s i f i c a t i o n : P r i m a r y 11L03, 11M99; Secondary 40D99.


K e y w o r d s : Riemann zeta function, finite sum.

149

DJURDJE CVIJOVIC JACEK KLINOWSKI

valid for any integer n > 2. For the case when n is even, he found the asymptotic
expansion of the finite sum involved, which readily leads to the Euler relation
n

1
,n-M.
C(2n) = (-l)"^eVB 2

(n>l)

2(2n)!'

between the even-indexed Bernoulli numbers B2n


Here, we deduce summation formulae for

(2)

and ((2n)

n + P K
~
Sn(q;0 = Y,

(0 < < 1)

(a)

P = i

and

-i

= Ecot"()
p=l

(b)

and examine the Apostol formula (1) for even integers.

2. First results
Throughout the text [x] denotes the largest integer not exceeding x, while,
as usual, (^) is the binomial coefficient given by
n\

and

,l

0/

f n\

n\
j -=-'
\mj
m\(n m)\

(n.mGN,

l<m<n).

Further, the principal values of arccot# are defined for any real 6 by 0 <
arccot x < 7T.
In this section we establish the existence of the polynomials for which all
zeros, regardless of the degree, can be given by a simple formula involving the
cotangent function. It appears that these polynomials are unknown, and their
detailed study is given elsewhere [3].
THEOREM. Let n be a positive integer, and Cn(x;) a polynomial of degree
precisely n. in a variable x with real parameter , which we define as
Cn(x;0=J2(-iyncn,m(0xn-m

(4a)

m=0
Where

n,m(0

is

ciTO(f) = Q
150

[cos(m7r/2) + sin(m7r/2) c o t ) ]

(0 < f < 1)

(4b)

FINITE COTANGENT SUMS AND THE RIEMANN ZETA FUNCTION

Then for any fixed n, the zeros of Cn(x\ ) are all real, simple and given by
r , m ( 0 = cot ( - i - t p - - - )

m = 0.1....,n-l.

(5)

Before proving the Theorem we give as an example the first few polynomials
Cx (x\ ) = x cot(7r),
C2 (~; f) = x2 - 2x c o t ) - 1,
C3(x] f) = x3 - 3x2 cot(7rf) - 3x -F COt(7Tf) ,
C4(x] f) = x4 - 4x 3 cot(7r0 - 6x2 + 4~cot(7rO + 1,
C5(x; f) = x5 - 5x 4 cot(?rO - 10x 3 + 10x 2 cot(Trf) + 5x - cot(Trf),
C6(x] ) = x 6 - 6x 5 cot(7r0 - 15x4 + 20x 3 c o t ^ ) + I5x2 - 6x cot(?rO - 1.
P r o o f . First of all, we shall prove that
[n/2]

V^(-l)m(
m=o
^

)xn~2m
m

= csc n (arccotx)cos(narccotx)

(6a)

and
[(n-l)/2]

U
Yl
(_1)m(
m=0
Vm +

) ^ n - 2 m _ 1 =cscn(arccotx)sin(narccotx)
/

(6b)

where arccot x takes its principal value, x is real and n is a positive integer.
Starting from the well-known Euler-Bernoulli formulae [4; 4.4.4.15, 4.4.4.16]
n/2]
["/-]

/
m

COS.
i6>=

](-l) (
m=0

'

[ ( n - l ) / 2 ]

sin?
m

)cosn-2msin2m>
,

Y
- =0
m

)cosn_2m-1sin2m+10

(-l) (

V2m+lj

we obtain
[n/2]
-L

Y2 ( - l ) m (

) tan2m

m=0
m
=0

'

= sec n

cosn

[(n-l)/2]

"

---1

mII

( - l ); (

]tan

V2m+11
N

2 m + 1

0 = sec n sinn

'

where 0 7^ (2fc+l)7r/2 (k Z ) . In view of these identities, upon setting x cot 0


(0 < 9 7^ 7r/2 < 7r) in (6a), (6b), it follows at once that the proposed formulae
151

DJURDJE CVIJOVIC JACEK KLINOWSKI


hold when x / 0. Moreover, it can be shown by direct verification that they
remain valid for x = 0.
Next, since the polynomials C n (x;) may be expressed in the form
[n/2]

[(n-l)/2]

2m

cn(^o=E^r(2my- -cot(,o
x

ra=0

'

ra=0

(-ir(2m+iy 2 m - l
v

(7)

as an immediate consequence of (6) we have


C n (x;) = csc n (arccotx)[cos(narccotx) - cot(7r)sin(narccotx)] .

(8)

To complete the proof we only need to verify that Cn(Tn m ();) = 0 for
any n and ra = 0 , 1 , . . . , n 1, regardless of .
Observe that, in general, if k is an integer and 9 is real, then
arccot(cot 9) = 9 - kir

kir < 9 < (k + 1)TT .

Thus

-*ccot[r n>m (0] =


since
0<

(!+mK<

(-r-rn)7r

m = 0 , 1 , . . . , n 1,

provided that 0 < < 1.


Finally, it follows readily from (8), that

c B ( r B i m ( 0 ; 0 = (-i) m c8c^ +

)''

[C0S(7) - COt(7) Sn(7<;)] = 0 .

Since Cn(x\^) is of degree n , it vanishes at the n points r n m ( ) for ra =


0 , 1 , . . . , n 1. In view of the properties of the cotangent functions, these zeros
are all real and simple. This completes our proof.

A similar result is obtained as follows. Let Cn(x) denote the polynomial given
by the expressions in (6b). Several examples are

C{(x = 1,

c;(x
c;(x

= 2x,
= x 2 - 1,

C*A(x = 4x 3 - 4x,
c;(x = 5x 4 - 10x 2 + 1,

c;(x
c;(x
c;(x

= 6x 5 - 20x 3 + 6x ,
= 7x6 - 35x 4 + 21x 2 - 1,

= Sx7 - 56x 5 + 56x 3 - 8x .


The determination of the zeros of Cn(x) is clearly equivalent to solving
sin(n arccot x) = 0, which gives rise to the following Proposition.
152

FINITE COTANGENT SUMS AND THE RIEMANN ZETA FUNCTION

PROPOSITION 1. Assume that n is a positive integer. Let Cn(x) be the polynomial in a real variable x of degree n 1. defined by (6b). Then for any fixed
n>2,
the zeros of Cn(x) are all real, simple and given by
r

;, = c o t ( ^ ) ,

m = l,2,...,n-l.

3. Finite cotangent sums


As a straightforward consequence of the Theorem, we derive a closed-form
determinant formula for finite cotangent sum Sn(q;). We note that the sums
Sx, S2 and S4 are well known in a slightly different form [4; Section 4.4.7] and
[5; Sections 29.1 and 30.1].
We begin by stating some elementary results from the theory of polynomials
needed [6; p. 179]. Let xk (1 < k < n) be the roots of an algebraic equation of
the form
xn + axxn~x
n

h an_xx +

\-an

n x

= x - axx - + + ( - l r - X - i * + ( - ! ) X =

(10)

with coefficients a1,a2,...


, a n (n > 1). The elementary symmetric functions
a1,a2,...ian
of xk are then given by the Viete formulae
x

i + x2 + * *' + xn = "i >

X^X2 + X1X^ + ' ' + "n\~n


X^2X^

+ XXX2X + ' ' ' +

<~2 5

x
X
X
n-2 n-l n

= a3 '

x1x2 + "- + xn = an
while sk (k = 1,2,... ), the kth power sum of zeros, defined as
sk = xl + x2 + - - + xn
is related to ak by the Newton identities as follows
"i =

sxax -2a2

i >

= 52,

s2al S-^a2 oO"3 = 5 3 ,

ki - sk-i(T2

+ + (-l)k+1kak

= sk .
153

DJURDJE CVIJOVIC

JACEK KLINOWSKI

The Newton identities can clearly be considered as a system of linear equations


for s-_, s 2 , . . . , sn, which enables us to rewrite them into a more convenient form.
Thus by making use of the Cramer rule [6; p. 198] it is easy to see from the first
k of the equations, that power sum sk is given by the following determinant
0
1

1
2a2

0"i

3<7o
Si.

(k

0
0
1

.. .
.. .
.. .

1
<,

c k-2

kau

0
0
0

k-1

of the order k.
Now, in view of the Theorem, it is obvious that the finite sum in (3a) is in
fact the power sum of the zeros of the polynomial in (4). Thus, we can state the
following result.
PROPOSITION 2. Assume that n is a positive integer, and let Sn(q;,)
finite sum in (3a). Then for any fixed n we have
c

2c9
C

sn(q;0
n-

q,2
<7,3

-K.n-1
nc
~q,n

where, for brevity, c


is of the order n.
In particular
Sx(q;0
S2(q;0
S3(q;0

<7,1

?,2

0
1
c
?

,l

0
0
1

.. .
..
.. .

,q,nq,n l

stands for c t(0

0
0
0
l
<7,1

defined in (4b), and the

determinant

qcot(nO,
2

= q (cot2(7r0
3

l)-q,

= <7 (cot (>r) + c o t ( O ) - g c o t f r O ,

SA(q; 0 = q4(cotA(7rO + (4/3) cot2(7r) + 1/3) - </2((4/3) c o t 2 ) + 4/3)


5

S5(q; 0 = q (cot (irO

+ (5/3) cot (7rO + (2/3) cotfrO)

- q ((5/3) cot 3 (7r0 + (5/3) cot 3 (7re)) + gcot(vrO ,


S6(q;0

= 9 6 (cot 6 (7r0 + 2cot 4 (7r) + (17/15) cot 2 (7rO + 2/15)


- q4 (2 cot 4 (7r0 + (8/3) cot2(7r) + 2/3)
+ q2 ((23/15) cot 2 (?r0 + 23/15) - q.

154

be a

+ q

FINITE COTANGENT SUMS AND THE RIEMANN ZETA FUNCTION

We note that the expressions for Sn(q\l/2),


Sn(q]l/4),
S2n(q;l/3)
and
S2n(q;l/6)
are the polynomials in q with rational coefficients.
Moreover,
___,(.; 1/2) = 0.
However, it is easier to find the sums for higher n by using the differential
recurrence relation

~W?;0 = -s-fofl - j^ihjnTzs^{q>


which is readily proved by differentiating

{n

-1}

the definition (3a) with respect to .

4. The Apostol formula


In this section we shall derive the closed-form formula for 5* (q) which allows
us to investigate the Apostol formula.
The properties of the cotangent function readily lead to
o-l

X>tW^W+l(?) = o
\ * /

P=I

and
o-l

X>|^
x

p=\

The first indication about S2n


Section 3 and the relation

'

can be obtained from the expressions found in

o-l

2n(?) = J E C O t " ( ^ 7 ^ ) = l[Sn(*0 ~ */_)]

resulting in
1 ,

'-^=3'

"

e*^
! 4 4 2__
26
54(9) = 45? - 9 ? + 9 - 4 5 .
2
o., .
2 4
23 2
502
6
(
) =
*
945^-45^ + 4 5 ^ - * + 9 4 5 '
where our S^ and 5 4 are the same as those listed in literature [4; Section 4.4.7].
5

155

DJURDJE CVIJOVIC JACEK KLINOWSKI

On the other hand, a straightforward consequence of Proposition 1 is that


the sum in (3b) is the power sum of the zeros of Cn(x). In this way we shall
evaluate Sn for any n. Note, however, that the polynomial Cn(x) is not monic,
that it is of degree n 1 and that some of its coefficients are zero (compare with
Equation 10).
PROPOSITION 3. Let n be a positive integer and let Sn(q)
defined by (3b). Then for any fixed n we have
1
0

0
2c9,2
0

S*n(q)
(n l)c v

where c

/ q,n l
q,n

0
1
0

,2

be a finite sum

0
0
1

cq,n

2
<
q,n-1

stands for c {(), and the determinant

is of the order n.

It is not difficult to verify that S^n-t-i ls a i w a v s zero, and it follows directly


that, in general, S%n is the polynomial of degree 2n in q with rational coefficients. Two more special cases are

S*s(q) =

1
4725

16
28354

"<") = d b ' " -

176

44

7102
?

+ Q

675
315
86
718
q +
q
2835
8505 ~ 8505

14175'

563
+

945

44834
93555

5. Concluding remarks
The Apostol formula implies the existence of a certain function A(q; n), such
that
A(q;n)

() = (

Cv

'

4 2/

li -ti-!l
n

q-+oo
?^oo

qqn

(n>2)

'

We have demonstrated that, for even integers, we have


A(2;2n) = - 5 * n ( 2 g + l )

(n>l),

where S^n is fully determined: it is a polynomial in q with rational coefficients


of degree 2n and its evaluation is straightforward.
156

FINITE COTANGENT SUMS AND THE RIEMANN ZETA FUNCTION

REFERENCES
[1] M A G N U S , W . O B E R G E T T I N G E R , F.SONI, R. P . : Formulas and Theorems for the
Special Functions of Mathematical Physics, Springer-Verlag, Berlin, 1966.
[2] A P O S T O L , T. M . : Another elementary proof of Euler's formula for (2), Amer. M a t h .
Monthly 80 (1973), 425-431.
[3] C V I J O V I C , D . K L I N O W S K I , J . : A new polynomial system with easily deducible zeros,
(Submitted).
[4] P R U D N I K O V , A. P . B R Y C H K O V , YU. A . M A R I C H E V , O. I. : Integrals and Series,
Vol. 1, Gordon and Breach Science Publishers, New York, 1986.
[5] HANSEN, E. R . : A Table of Series and Products, Prentice-Hall, Englewood Cliffs, 1975.
[6] C O H N , P . M . : Algebra, Vol. 1, John Wiley, Chichester, 1982.

Received January 26, 1998

Department of Chemistry
University of Cambridge
Lensfield Road
Cambridge CB2 1EW
UNITED
KINGDOM

157